WAV adaptor processing

Voice communications in WAV format are processed through different adaptors, Kafka processes, and Streams jobs.

IBM® Surveillance Insight® for Financial Services processes WAV files based on the metadata trigger that is received through pre-defined Kafka topics. The WAV adaptor reads the data from the Kafka topic, decrypts the Kafka message, parses it, and fetches the voice audio file location. The audio content is then passed to the SpeechToText (S2T) toolkit operator for translation. All of the utterances and the speaker diarization are aggregated. The aggregated conversation text is then converted to a communication object and then published to the Kafka topic.

Also, if an export URL is configured, the voice artifacts—the metadata, utterances, and the audio binary—are sent to the export service.

The export capability allows you to export individual voice artifacts to different endpoints. You can specify the following parameters when you submit the Streams job:

To export all of the voice-related artifacts to the HDFS on hostname1:

EXPORTALLURL=https://<hostname1>:<port>/SIFSVoiceDataService/voice/v1/export

To export the voice metadata-related artifacts to the HDFS on hostname2:

EXPORTMETADATAURL=https://<hostname2>:<port>/SIFSVoiceDataService/voice/v1/export

To export the voice metadata and transcript to the HDFS on hostname3:

HDFS.EXPORTTRANSCRIPTURL=https://<hostname3>:<port>/SIFSVoiceDataService/voice/v1/export

To export the voice metadata and audio data to the HDFS on hostname2 and hostname4:

EXPORTAUDIOURL=https://<host2>:<port>/SIFSVoiceDataService/voice/v1/export;https://<hostname4>:<port>/SIFSVoiceDataService/voice/v1/export

Diagram showing the WAV file processing — Figure 1. WAV